Early Human Development, 24 (1990) 31-45 Elsevier Scientific Publishers Ireland Ltd.
31
EHD 01087
Normal fetal growth evaluated by longitudinal ultrasound examinations Torben Larsenalb, Sten Petersenc, Gorm Greisenc and Jbrgen Falck Larsenb ‘Department of Ultrasound, Wepartment of Obstetrics and Gynaecology, Herlev Hospital and TIepartment of Paediatrics, Glostrup Hospital, University of Copenhagen, Copenhagen (Denmark) (Received 15 January 1990; revision received 25 April 1990; accepted 4 July 1990)
Summary Fetal weight estimation was evaluated using the equations of Warsof, Shepard and Hadlock in 192 patients, less than 3 days before delivery. Warsof’s and Hadlock’s equations resulted in significantly better weight estimates compared to Shepard’s equation. No systematic error was found below 2500 g by use of Warsof’s equation, whereas Shepard’s and.Hadlocks’s equations resulted in significant overestimation in the low weight group. In a study of 5 fetuses, of 27-38 weeks gestational age, the intra-observer variation was calculated to 4.6070, whereas the coefficient of variation among observer means was 2.9%. The mixed intra- and inter-observer coefficient of variation was 6.5%. Thirty-five low-risk, uncomplicated pregnancies with reliable last menstrual dates were investigated longitudinally with ultrasound measurements of fetal weight. Population growth curves of fetal weight, fetal femur length, abdominal circumference and biparietal diameter were constructed by weighted polynomial regression. After 27 weeks of gestational age the weight growth curve showed only insignificant non-linearity. Compared to a Danish growth curve based on birth weights, significant higher mean weight was found, especially before 31 weeks of gestational age. The 10th and 90th percentiles for the individual percentage deviation change was + 4.4% per 28 days. fetal growth; ultrasound.
Correspondence to: Torben Larsen, Department of Obstetrics and Gynaecology, Herlev Hospital, DK2730 Herlev, Denmark. 0378-3782/90/$03.50 0 1990 Elsevier Scientific Publishers Ireland Ltd. Published and Printed in Ireland
Introduction Perinatal growth charts are usually based on birth weight in term and preterm infants [1,2]. However, many obstetrical risk factors associated with preterm birth are also associated with intrauterine growth retardation (IUGR) [3,4]. Therefore it might be suspected that infants born preterm may have lower body weight than those of the same gestation remaining in utero. This is supported by the observation that growth charts based on data from selected low-risk pregnancies show higher weight for gestational age (GA) than those based on the general population [5]. Pregnancies ending in preterm deliveries, however, can never be considered normal. Therefore intrauterine growth can only be evaluated from measurements of size of fetuses in utero. Such growth charts based on ultrasound have been presented. However, significantly different mean weights have been found [6-81. Individual fetuses do not necessarily follow the population growth curve. The aim of the present study was to establish normal longitudinal fetal growth, with GA based on early ultrasound (US) measurement of the biparietal diameter (BPD), in order to use the results in a prospective longitudinal study of risk pregnancies. Methods Measurement procedures All measurements were carried out by four sonographers and one of the authors (T.L.). A linear array scanner with sound velocity calibrated to 1540 m/s was used. Electronic calipers were used for all measurements. The median of three measurements was recorded. BPD was measured at right angles to the longitudinal axis of the skull with a clear midline echo and clearly discernible lateral ventricles. BPD was measured from the outer edge of the anterior skull wall to the inner edge of the posterior skull wall. Abdominal circumference (AC) measurement was performed at right angles to the fetal aorta or spine at the level where the umbilical vein could be clearly visualized within a “circular” abdomen. The antero-poster0 and an orthogonal diameter were measured from outer limitation of the circumference. The geometric mean (AD) was used to calculate AC, from AC = AD * II. Femur length (FL) was measured from the greater trochanter to the lateral condyle. Evaluation of three fetal weight estimation equations In 192 risk pregnancies examined by US less than 3 days before birth, estimated fetal weight (eFW) using Warsof’s [9], Shepard’s [lo] and Hadlock’s [l 11equations (Table I) were compared with birth weight. The interval between US examination and birth was less than 1 day in 56 cases, 1-2 days in 75, and 2-3 days in 61 of the patients. Before analysis, birth weights were corrected with the assumed average weight gain of 28 g/day from US examination to birth. The Warsof equation resulted in significantly better weight estimates for the
39
TABLE I Percentage deviation of fetal weight estimates using the equations of Warsof’, Shepardb, and Hadlock’ from birth weights corrected with 28 g/day in different weight groups. Corrected BW
n
All *2500 2501-4000 >4000
192 34 146 12
Warsof Mean (olo)
S.D. (olo)
Shepard Mean (@/o) SD. (To)
Hadlock Mean (070)
S.D. (Ore)
- 2.2*** 1.50 - 2.6**+ -6.1*
8.7 10.6 8.1 7.9
2.6*** 7.0** 2.1** -2.3’=
2.0.’ 5.8’ 1.72’ - 3.80
8.3 11.0 7.4 8.8
9.1 11.1 8.5 8.0
’Warsof’s equation (eFW in kg, BPD and AC in mm): log,, (eFW) = - 1.599 + 0.0144 * BPD + 0.0032 -ACl.llE--7.BPD.BPD.AC. bShepard’s equation (eFW in kg, BPD and AC in mm): log,0 (eFW) = - 1.7492 + 0.0166 - BPD + 0.0046 - AC - 2.646E-5 - BPD - AC. EHadlock’s equation (eFW in gram, BPD, AC and FL in mm): log,, (eFW) = 1.335 - 3.4E-5 * AC * FL + 0.00316 - BPD + 0.00457 - AC + 0.01623 - FL. t-test: %.s., *P< 0.05, ?? *P< 0.01, ***P< 0.001.
group as a whole than did Shepard’s equation (P < O.OOl), but was not different from that obtained by using Hadlock’s equation. However, below 2500 g the estimated weight using Warsof’s equation was not significantly different from the corrected birth weight (Table I). Using Warsof’s equation, 95 infants (49%) had corrected birth weights within + 5% and 140 infants (74%) within f 10% from the estimated fetal weights. Repeatability study This study provided 9 estimates of fetal weight, 3 by each of 3 observers, for 5 fetuses, of 27-38 weeks GA. Electronic calipers were used for measurements, but the results were blinded on the monitor. Results were documented on Polaroid photographs. Data were analyzed by two-way analysis of variance. In each fetus the individual weight estimates were normalised by their mean. The test-retest coefficient of variation for the Warsof weight estimate (same observer, same fetus) was 4.6%. The coefficient of variation among observer means (inter-observer variation) was 2.9% (0.05 < P < 0.1). The mixed inter- and intraobserver coefficient of variation (calculated from the overall mean square) was 6.5%. The coefficient of variation within fetuses was not related to the mean estimated weight of the fetus (P > 0.2), supporting the notion of constant coefficient of variation. For BPD, AD and FL the test-retest variation was 0.95 mm, 2.33 mm and 2.35 mm, respectively. For none of the measures was the inter-observer variance significantly different from zero. Normal material Forty-nine low-risk patients, with known date of last menstrual period, regular
40
menses (duration 4-7 days, intervals of 28 2 4 days), participated in a longitudinal study including fetal ultrasound measurements every two weeks from 15 weeks of gestation. Four of these patients were subsequently excluded as their pregnancies diverged from normal; two because of abruptio placentae, one because of preeTABLE II Biparietal diameter (BPD), femur length (FL), abdominal circumference (AC), and estimated fetal weight (eFW) observed for each completed gestational week. BPD (mm)
FL (mm)
AC (mm)
eFW (9)
geks) Mean
SD.
Mean
S.D.
Mean
14 15
29 33
0.7 1.1
15 17
1.3 1.9
95 103
6.5 7.7
16 17 18 19
36 39 42 46
20 23 26 29
24 25 26 27
48 51 55 58 61 64 68 70
1.5 1.7 2.6 2.5 1.5 2.1 2.5 2.3 2.6 2.5 1.8 3.2
112 126 137 148
20 21 22 23
1.2 1.5 1.9 1.9 1.9 2.0 1.5 2.2
28 29 30 31
74 76 78 82
2.9 3.5 2.3 3.5
32 33 34 35
83 84 86 88
36 37 38 39
90 91 92 93
3.5 3.6 3.6 2.8 3.2 4.1 2.7 3.9
53 54 57 59 61 62 66 66 69 69 72 72
7.9 6.8 7.5 9.6 7.4 10.4 7.4 10.2 12.7 11.8 9.9 10.1 11.7 12.2 13.4 12.6
40
96
3.9
76
Regression
1.7 2.6 2.6 2.7
31 34 38 40 42 45 48 50
3.0 2.8 2.4 2.2 2.6 2.3 2.7 2.6 3.9 2.5 2.2 3.2 2.9
SD.
157 169 185 190 206 214 228 238 247 255 269 278 291 300 314 322 329 338 343 349
13.5 12.1 13.9 12.5 15.5 14.8 16.8 20.6
369
14.5
Mean
S.D.
347 376 432 525 591 719 811 981 1102 1282 1399 1595 1841 2015 2160 2454 2645 2866 3066 3222 3436 3831
19.8 16.7 39.4 33.9 57.6 71.4 90.2 89.1 126.6 150.7 185.0 158.1 234.2 250.9 264.0 299.2 262.3 358.1 399.5 356.9 344.4 394.7
lines
Biparietal diameter: BPD = -22.4 + 0.5126 * GA - 4.35371E-9 * GA’, R2 = 0.99, SEE = 1.53, Durbin-Watson = 0.62. - GA’, R’ = 0.98, SEE = 1.56, DurFemur length: FL = - 31.% + 0.47 - GA - l.l%534E-6 bin-Watson = 1.11. Abdominal circumference: AC = - 81.0 + 1.70 * GA - 566994E-9 * GA’, R* = 0.98, SEE = 11.7, Durbin-Watson = 1.10. Fetal weight: FW = 2315.6 - 26.46 -GA + 8.79618E-4 - GA’ - 1.7267956E--6. GA’, R’ = 0.97, SEE = 13.3, Durbin-Watson = 0.69.
41
clampsia and one because her smoking habits eventually exceeded 15 cigarettes a day. Ten other patients discontinued participation in the study, leaving for evaluation 35 patients with 4 or more examinations after day 133 (19 completed weeks). Eleven patients were nulliparae, 18 were primiparae and 6 patients were multiparae. All patients delivered spontaneously. GA was calculated from the first BPD measurement [12]. Fetal weight estimates were those based on Warsof’s equation [9]. Population growth curves were estimated from all available data by use of weighted polynomial regression (up to the sixth order). The weighting, as a function of GA, was taken as the reciprocal of the standard deviation for each gestational week (Table II). Linear approximation of sections of the fetal weight growth curve were tested by comparison of residuals by the F-test. Three types of individual growth curves from day 189 of gestation until birth were fitted to the fetal weight as percentage deviation from the population curve: (a) constant percentage deviation, (b) linear change of percentage deviation and (c) 2nd order, allowing a deceleration-acceleration or acceleration-deceleration in the period under study. The mean squared residuals of the three models were compared by the F-test. Finally, the mean squared deviation of predicted BW from actual BW, using the three different growth models, were compared, also by the F-test. Informed consent was obtained before participation and the study was approved by the local ethics committee. Results The 35 low-risk women included, gave birth to 22 boys and 13 girls. Mean gestational age was 282 days (S.D. 8.4 days) which is not significantly different from 280 days. Mean gestational age calculated from first day of last menstrual period was 285 days (S.D. 9.5 days) which is significantly different from 280 days (P < 0.005). Mean birth weight was 3660 g (S.D. 560 g), mean crown-heel length was 524 mm (S.D. 27 mm) and mean head circumference was 353 mm (S.D. 15 mm). The longitudinal study of the 35 pregnancies comprised 303 fetal weight estimations from day 133 until birth (Fig. 1). The raw data and the equations of the regression lines are given in Table II. For none of the fetal dimensions measured, were polynomials of more than 4th order necessary to obtain the best fit. After restricting the analysis to the period from day 189 (27 weeks) and until birth, leaving 192 data sets, the weight growth curve showed only insignificant nonlinearity(F = 2.18, d.f. = 2,189, P > 0.1). The linear approximation to the ‘population’ growth curve was: fetal weight (FW) = (GA - 154) * 28.04. The coefficient of variation was 11.4%. The difference in mean weights for girls and boys was 3.8% in this material. In the calculation of lOth, 50th and 90th percentiles for girls and boys (Table III), the data were corrected for a weight differences between girls and boys of 4%. Fitting individual growth curves from 27 weeks until birth, allowing each fetus its own constant percentage deviation (growth channel), reduced the mean square very
42
4000
1
3500. ,3000z2500r E2000s
15001000 500
1
1z 04 133
154
175
i96
217
238
259
280
1
301
Gestational age (days) Fig. 1. Estimated fetal weight in 35 normal pregnancies. The linear population growth curve from 27 weeks until term is indicated by its 10th. 50th and 90th percentiles.
TABLE III Fetal weight percentiles for girls and boys. Gestational age
Girls
Weeks
Days
10%
27 28
189 1%
817 981
956 1148
29 30 31 32
203 210 217 224 231 238 245 252
1144 1308 1471 1635 1798 1%2 2125 2289
1339 1530 1722 1913 2104 22% 2487 2678
259 266 273 280 287 294
2452 2616 2179 2943 3106 3270
2870 3061 3252 3444 3635 3826
33 34 35 36 37 38 39 40 41 42
Boys 50%
90070
1095 1315 1534 1753 1972 2191 2410 2630 2849 3068 3287 3506 3725 3945 4164 4383
10%
50%
90%
851 1021 1191 1361 1532 1702 1872 2042 2213 2383
995 1195 1394 1593 1792 1991 2191 2390 2589 2788
2553 2723 2893 3064 3234 3404
2987 3187 3386 3585 3784 3983
1140 1368 1597 1825 2053 2281 2509 2737 2966 3194 3422 3650 3878 4106 4335 4563
43
significantly (F = 23.81, d.f. = 34, 157, P 3 O.OOl), leaving a coefficient of variation of 5.3%. The individual growth channels were normally distributed with standard deviation 10.2%. The 10th and 90th percentiles were calculated to f 13.0%, corresponding to a daily weight gain of 24.2 g/day and 31.6 g/day, respectively. Allowing constantly changing percentage weight deviation further improved the fit (F = 1.77, d.f. = 35,122, P< 0.025) and reduced the coefficient of variation around the individual growth curves to 4.9% which is of the same magnitude as the test-retest coefficient of variation. The 10th and 90th percentiles for the individual percentage deviation change was f 4.4% per 28 days. Using 2nd order non-linear individual growth curves did not improve the fit further. The mean deviation of actual birth weight from FW at the date of birth predicted from individual constant deviation growth curves was 7.2%. This deviation was not decreased by allowing changing percentage deviation: 8.6% (P > 0.05) and was similar to that found when fetal weight estimation was within 72 h of birth (Table I). Discussion In our material the divergence of estimated fetal weight from actual birth weight was acceptably low ( - 2.2%) when the equation proposed by Warsof et al. [9] was used. Using Shepard’s [lo] and Hadlock’s [l l] equations we found a significant overestimation of 7.0% and 5.8070, respectively, in the weight group below 2500 g. For diagnosing IUGR, correct fetal weight estimation in the weight group below 2500 g is particularly important and therefore we have chosen fetal weight estimation using Warsof’s equation in establishing our reference material. The correction of birth weight with 28 g per day interval between US and birth tends to decrease underestimation and increase overestimation. Longitudinal studies of severely growth retarded fetuses, however, show that daily growth of less than 15 g from 27 week is rare. Even an average overcompensation of 13 g per day, however, will not influence our conclusion in the comparison of the three weight estimation equations. For all types of growth charts the establishment of GA is essential. Since early BPD measurements are superior to menstrual data in calculating term [ 131and since routine early BPD measurements form the basis for clinical practice, it seems reasonable to use the same method for determination of GA in the construction of fetal growth curves. The data obtained from this study are suitable for comparison of growth deviation. The assumption of this statement is that the biologic variance of BPD before 20 weeks of GA results in only negligible variance in calculated terms. It is not reasonable of course, to determine GA from either the BPD, AC, or FL curves from this study since the start of our curves is determined by the BPD curve of Persson et al. [ 121. The normal fetal weight curve demonstrated significantly higher mean weights (about 5%) than a Danish growth curve based on cross-sectional analysis of birth weights (GA calculated from BPD) presented by Secher et al. [2] (Table IV). We found increasing discrepancy particularly below 3 1 weeks of GA. The small number of observations below 31 weeks in that study, and the probability of concomitant
44
TABLE IV Fetal weight from polynomial growth curve corrected for the number of boys and girls compared Secher’s data from liveborn infants and Persson’s data from fetal weights by ultrasound. Gestational
age
FW
Secher
Persson
S-FW
P-FW
(g)
(g)
(8)
(Qra)
(%o)
- 19.1 - 13.2 - 9.4 -7.1 - 5.7 - 5.0 - 4.8 -4.8 -5.0 - 5.2 - 5.4 - 5.6 - 5.1 - 5.6 -5.3 -4.8
8.7 7.5
Weeks
Day
21 28
189 196
1044 1198
845 1040
1135 1288
29 30 31 32
203 210 217 224
1364 1540 1725 1918
1236 1431 1626 1822
33 34 35 36 31 38 39 40
231 238 245 252 259 266 273 280
2118 2324 2534 2746 2959 3171 3380 3584
41 42
287 294
3780 3%7
2017 2213 2408 2603 2199 2994 3189 3385 3580 3776
1449 1617 1792 1973 2160 2351 2548 2749 2955 3163 3375 3589 3806 4024
with
6.2 5.0 3.9 2.9 2.0 1.2 0.6 0.1 0.1 0.3 0.1 0.1 0.7 1.4
growth retardation in very preterm infants seems likely to be responsible for the discrepancy. The daily growth rate of 28 g in our study was similar to that found by Secher, as was the absence of flattening in the growth curve towards term. It is likely that a more precise estimation of GA from early BPD measurement is the main reason for this. Comparing our results with those given by Persson et al. [8] in a similar study of 19 pregnancies, but using their own weight estimating equation, we found identical mean values from 30 to 40 weeks, but lower values below 31 weeks (Table IV). Deter et al. [6] presented estimates in a longitudinal study of 20 patients, using Warsof’s equation. Our results are 1.5% lower in the range of 30 to 34 weeks of GA. Also our values below and above this interval are increasingly different. It seems likely that the differences especially in the early and late gestational weeks are results of the mathematical modelling rather than of differences in population or in the technique used for biometric measurements. The coefficient of variation of the method of estimating weight from BPD and AC may be calculated from the mixed inter- and intra-observer coefficient of variation of 6.5% (i.e., reproducibility) and the mean coefficient of variation of the birth weight from FW of 8.7%: \/(8.72 - 6J2) = 5.8% (i.e. reliability: the difference between actual FW and the mean value of a large number of ultrasound weight estimates of that fetus). However, it can not be expected that individual body weight will be precisely reflected by only two or three linear body dimensions. On the other
45
hand, as there was no evidence of systematic error in the weight estimate, it is possible to use the calculated means as a measure of true mean FW. In our longitudinal study of FW the 10th and 90th percentiles for individual changes in growth rate (acceleration or deceleration) was f 4.4% per 28 days. When IUGR is diagnosed from longitudinal estimation of FW this limit of normal variations in growth rate is important since an abnormal drop in weight percentile might be a more precise indication of IUGR than a single observation of FW being below the 10th percentile. Acknowledgements The study was supported by grants from Ville Heises Legat (26/89), Sygekassernes Helsefond (37/89), Dronning Louises B$rnehospitals Forskningsfond (17/85 and 16186) and Lundbeckfonden (147188). References 1 2
3
4 5 6
I 8 9 10
11
12
13
Lubchenco, L.O., Hansman, C., Dressler, M. and Boyd, E. (1963): Intrauterine growth as estimated from liveborn birth-weight data at 24 to 42 weeks of gestation. Pediatrics, 32,793-800. Secher, NJ., Kern Hansen, P., Lenstrup, C., Pedersen-Bjergaard, L. and Sindberg Eriksen, P. (1986): Birthweight-for-gestational age charts based on early ultrasound estimation of gestational age. Br. J. Obstet. Gynaecol., 93, 128-34. Kaminski, M. and Papiernik, E. (1974): Multifactorial study of the risk of prematurity at 32 weeks of gestation. II. A comparison between an empirical prediction and a discriminant analysis. J. Perinat. Med., 2,37-44. Weidinger, H. and Wiest, W. (1974): A comparative study of the epidemiological data of pregnancies with and without tendencies to premature delivery. J. Perinat. Med., 2,276-87. Ulrich, M. (1982): Fetal growth patterns in a population of Danish newborn infants. Acta Paediatr Stand. Suppl. 292. Deter, R.L., Harrist, R.B., Hadlock, F.P. and Poindexter, A.N. (1982): Longitudinal studies of fetal growth with the use of dynamic image ultrasonography. Am. J. Obstet. Gynecol., 143, 54% 554. Jeanty, P., Cantraine, F.. Romero, R., Cousaert, E. and Hobbins, J.C. (1984): A longitudinal study of fetal weight growth. J. Ultrasound Med., 3,321-328. Persson, P.-H. and Weldner, B.-M. (1986): Intra-uterine weight curves obtained by ultrasound. Acta Obstet. Gynecol. Stand., 65, 169-173. Warsof, S.L., Gohari, P., Berkowitz, R.L. and Hobbins, J.C. (1977): The estimation of fetal weight by computer-assisted analysis. Am. J. Obstet. Gynecol., 128,881-892. Shepard, M.J., Richards, V.A., Berkowitz, R.L., Warsof, S.L. and Hobbins, J.C. (1982): An evaluation of two equations for predicting fetal weight by ultrasound. Am. J. Obstet. Gynecol., 142,47 -54. Hadlock F.P., Harrist R.B., Sharman R.S., Deter R.L. and Park, S.K. (1985): Estimation of fetal weight with the use of head, body and femur measurements - A prospective study. Am. J. Obstet. Gynecol., 151.333-337. Persson, P.-H. and Weldner, B.-M. (1986): Normal range growth curves for fetal biparietal diameter, occipito frontal diameter, mean abdominal diameters and femur length. Acta Obstet. Gynecol. Stand., 65,759-761. Campbell, S., Warsof, S.L., Little, D. and Cooper, D.J. (1985): Routine ultrasound screening for the prediction of gestational age. Obstet. Gynecol., 65,613-620.
31
EHD 01087
Normal fetal growth evaluated by longitudinal ultrasound examinations Torben Larsenalb, Sten Petersenc, Gorm Greisenc and Jbrgen Falck Larsenb ‘Department of Ultrasound, Wepartment of Obstetrics and Gynaecology, Herlev Hospital and TIepartment of Paediatrics, Glostrup Hospital, University of Copenhagen, Copenhagen (Denmark) (Received 15 January 1990; revision received 25 April 1990; accepted 4 July 1990)
Summary Fetal weight estimation was evaluated using the equations of Warsof, Shepard and Hadlock in 192 patients, less than 3 days before delivery. Warsof’s and Hadlock’s equations resulted in significantly better weight estimates compared to Shepard’s equation. No systematic error was found below 2500 g by use of Warsof’s equation, whereas Shepard’s and.Hadlocks’s equations resulted in significant overestimation in the low weight group. In a study of 5 fetuses, of 27-38 weeks gestational age, the intra-observer variation was calculated to 4.6070, whereas the coefficient of variation among observer means was 2.9%. The mixed intra- and inter-observer coefficient of variation was 6.5%. Thirty-five low-risk, uncomplicated pregnancies with reliable last menstrual dates were investigated longitudinally with ultrasound measurements of fetal weight. Population growth curves of fetal weight, fetal femur length, abdominal circumference and biparietal diameter were constructed by weighted polynomial regression. After 27 weeks of gestational age the weight growth curve showed only insignificant non-linearity. Compared to a Danish growth curve based on birth weights, significant higher mean weight was found, especially before 31 weeks of gestational age. The 10th and 90th percentiles for the individual percentage deviation change was + 4.4% per 28 days. fetal growth; ultrasound.
Correspondence to: Torben Larsen, Department of Obstetrics and Gynaecology, Herlev Hospital, DK2730 Herlev, Denmark. 0378-3782/90/$03.50 0 1990 Elsevier Scientific Publishers Ireland Ltd. Published and Printed in Ireland
Introduction Perinatal growth charts are usually based on birth weight in term and preterm infants [1,2]. However, many obstetrical risk factors associated with preterm birth are also associated with intrauterine growth retardation (IUGR) [3,4]. Therefore it might be suspected that infants born preterm may have lower body weight than those of the same gestation remaining in utero. This is supported by the observation that growth charts based on data from selected low-risk pregnancies show higher weight for gestational age (GA) than those based on the general population [5]. Pregnancies ending in preterm deliveries, however, can never be considered normal. Therefore intrauterine growth can only be evaluated from measurements of size of fetuses in utero. Such growth charts based on ultrasound have been presented. However, significantly different mean weights have been found [6-81. Individual fetuses do not necessarily follow the population growth curve. The aim of the present study was to establish normal longitudinal fetal growth, with GA based on early ultrasound (US) measurement of the biparietal diameter (BPD), in order to use the results in a prospective longitudinal study of risk pregnancies. Methods Measurement procedures All measurements were carried out by four sonographers and one of the authors (T.L.). A linear array scanner with sound velocity calibrated to 1540 m/s was used. Electronic calipers were used for all measurements. The median of three measurements was recorded. BPD was measured at right angles to the longitudinal axis of the skull with a clear midline echo and clearly discernible lateral ventricles. BPD was measured from the outer edge of the anterior skull wall to the inner edge of the posterior skull wall. Abdominal circumference (AC) measurement was performed at right angles to the fetal aorta or spine at the level where the umbilical vein could be clearly visualized within a “circular” abdomen. The antero-poster0 and an orthogonal diameter were measured from outer limitation of the circumference. The geometric mean (AD) was used to calculate AC, from AC = AD * II. Femur length (FL) was measured from the greater trochanter to the lateral condyle. Evaluation of three fetal weight estimation equations In 192 risk pregnancies examined by US less than 3 days before birth, estimated fetal weight (eFW) using Warsof’s [9], Shepard’s [lo] and Hadlock’s [l 11equations (Table I) were compared with birth weight. The interval between US examination and birth was less than 1 day in 56 cases, 1-2 days in 75, and 2-3 days in 61 of the patients. Before analysis, birth weights were corrected with the assumed average weight gain of 28 g/day from US examination to birth. The Warsof equation resulted in significantly better weight estimates for the
39
TABLE I Percentage deviation of fetal weight estimates using the equations of Warsof’, Shepardb, and Hadlock’ from birth weights corrected with 28 g/day in different weight groups. Corrected BW
n
All *2500 2501-4000 >4000
192 34 146 12
Warsof Mean (olo)
S.D. (olo)
Shepard Mean (@/o) SD. (To)
Hadlock Mean (070)
S.D. (Ore)
- 2.2*** 1.50 - 2.6**+ -6.1*
8.7 10.6 8.1 7.9
2.6*** 7.0** 2.1** -2.3’=
2.0.’ 5.8’ 1.72’ - 3.80
8.3 11.0 7.4 8.8
9.1 11.1 8.5 8.0
’Warsof’s equation (eFW in kg, BPD and AC in mm): log,, (eFW) = - 1.599 + 0.0144 * BPD + 0.0032 -ACl.llE--7.BPD.BPD.AC. bShepard’s equation (eFW in kg, BPD and AC in mm): log,0 (eFW) = - 1.7492 + 0.0166 - BPD + 0.0046 - AC - 2.646E-5 - BPD - AC. EHadlock’s equation (eFW in gram, BPD, AC and FL in mm): log,, (eFW) = 1.335 - 3.4E-5 * AC * FL + 0.00316 - BPD + 0.00457 - AC + 0.01623 - FL. t-test: %.s., *P< 0.05, ?? *P< 0.01, ***P< 0.001.
group as a whole than did Shepard’s equation (P < O.OOl), but was not different from that obtained by using Hadlock’s equation. However, below 2500 g the estimated weight using Warsof’s equation was not significantly different from the corrected birth weight (Table I). Using Warsof’s equation, 95 infants (49%) had corrected birth weights within + 5% and 140 infants (74%) within f 10% from the estimated fetal weights. Repeatability study This study provided 9 estimates of fetal weight, 3 by each of 3 observers, for 5 fetuses, of 27-38 weeks GA. Electronic calipers were used for measurements, but the results were blinded on the monitor. Results were documented on Polaroid photographs. Data were analyzed by two-way analysis of variance. In each fetus the individual weight estimates were normalised by their mean. The test-retest coefficient of variation for the Warsof weight estimate (same observer, same fetus) was 4.6%. The coefficient of variation among observer means (inter-observer variation) was 2.9% (0.05 < P < 0.1). The mixed inter- and intraobserver coefficient of variation (calculated from the overall mean square) was 6.5%. The coefficient of variation within fetuses was not related to the mean estimated weight of the fetus (P > 0.2), supporting the notion of constant coefficient of variation. For BPD, AD and FL the test-retest variation was 0.95 mm, 2.33 mm and 2.35 mm, respectively. For none of the measures was the inter-observer variance significantly different from zero. Normal material Forty-nine low-risk patients, with known date of last menstrual period, regular
40
menses (duration 4-7 days, intervals of 28 2 4 days), participated in a longitudinal study including fetal ultrasound measurements every two weeks from 15 weeks of gestation. Four of these patients were subsequently excluded as their pregnancies diverged from normal; two because of abruptio placentae, one because of preeTABLE II Biparietal diameter (BPD), femur length (FL), abdominal circumference (AC), and estimated fetal weight (eFW) observed for each completed gestational week. BPD (mm)
FL (mm)
AC (mm)
eFW (9)
geks) Mean
SD.
Mean
S.D.
Mean
14 15
29 33
0.7 1.1
15 17
1.3 1.9
95 103
6.5 7.7
16 17 18 19
36 39 42 46
20 23 26 29
24 25 26 27
48 51 55 58 61 64 68 70
1.5 1.7 2.6 2.5 1.5 2.1 2.5 2.3 2.6 2.5 1.8 3.2
112 126 137 148
20 21 22 23
1.2 1.5 1.9 1.9 1.9 2.0 1.5 2.2
28 29 30 31
74 76 78 82
2.9 3.5 2.3 3.5
32 33 34 35
83 84 86 88
36 37 38 39
90 91 92 93
3.5 3.6 3.6 2.8 3.2 4.1 2.7 3.9
53 54 57 59 61 62 66 66 69 69 72 72
7.9 6.8 7.5 9.6 7.4 10.4 7.4 10.2 12.7 11.8 9.9 10.1 11.7 12.2 13.4 12.6
40
96
3.9
76
Regression
1.7 2.6 2.6 2.7
31 34 38 40 42 45 48 50
3.0 2.8 2.4 2.2 2.6 2.3 2.7 2.6 3.9 2.5 2.2 3.2 2.9
SD.
157 169 185 190 206 214 228 238 247 255 269 278 291 300 314 322 329 338 343 349
13.5 12.1 13.9 12.5 15.5 14.8 16.8 20.6
369
14.5
Mean
S.D.
347 376 432 525 591 719 811 981 1102 1282 1399 1595 1841 2015 2160 2454 2645 2866 3066 3222 3436 3831
19.8 16.7 39.4 33.9 57.6 71.4 90.2 89.1 126.6 150.7 185.0 158.1 234.2 250.9 264.0 299.2 262.3 358.1 399.5 356.9 344.4 394.7
lines
Biparietal diameter: BPD = -22.4 + 0.5126 * GA - 4.35371E-9 * GA’, R2 = 0.99, SEE = 1.53, Durbin-Watson = 0.62. - GA’, R’ = 0.98, SEE = 1.56, DurFemur length: FL = - 31.% + 0.47 - GA - l.l%534E-6 bin-Watson = 1.11. Abdominal circumference: AC = - 81.0 + 1.70 * GA - 566994E-9 * GA’, R* = 0.98, SEE = 11.7, Durbin-Watson = 1.10. Fetal weight: FW = 2315.6 - 26.46 -GA + 8.79618E-4 - GA’ - 1.7267956E--6. GA’, R’ = 0.97, SEE = 13.3, Durbin-Watson = 0.69.
41
clampsia and one because her smoking habits eventually exceeded 15 cigarettes a day. Ten other patients discontinued participation in the study, leaving for evaluation 35 patients with 4 or more examinations after day 133 (19 completed weeks). Eleven patients were nulliparae, 18 were primiparae and 6 patients were multiparae. All patients delivered spontaneously. GA was calculated from the first BPD measurement [12]. Fetal weight estimates were those based on Warsof’s equation [9]. Population growth curves were estimated from all available data by use of weighted polynomial regression (up to the sixth order). The weighting, as a function of GA, was taken as the reciprocal of the standard deviation for each gestational week (Table II). Linear approximation of sections of the fetal weight growth curve were tested by comparison of residuals by the F-test. Three types of individual growth curves from day 189 of gestation until birth were fitted to the fetal weight as percentage deviation from the population curve: (a) constant percentage deviation, (b) linear change of percentage deviation and (c) 2nd order, allowing a deceleration-acceleration or acceleration-deceleration in the period under study. The mean squared residuals of the three models were compared by the F-test. Finally, the mean squared deviation of predicted BW from actual BW, using the three different growth models, were compared, also by the F-test. Informed consent was obtained before participation and the study was approved by the local ethics committee. Results The 35 low-risk women included, gave birth to 22 boys and 13 girls. Mean gestational age was 282 days (S.D. 8.4 days) which is not significantly different from 280 days. Mean gestational age calculated from first day of last menstrual period was 285 days (S.D. 9.5 days) which is significantly different from 280 days (P < 0.005). Mean birth weight was 3660 g (S.D. 560 g), mean crown-heel length was 524 mm (S.D. 27 mm) and mean head circumference was 353 mm (S.D. 15 mm). The longitudinal study of the 35 pregnancies comprised 303 fetal weight estimations from day 133 until birth (Fig. 1). The raw data and the equations of the regression lines are given in Table II. For none of the fetal dimensions measured, were polynomials of more than 4th order necessary to obtain the best fit. After restricting the analysis to the period from day 189 (27 weeks) and until birth, leaving 192 data sets, the weight growth curve showed only insignificant nonlinearity(F = 2.18, d.f. = 2,189, P > 0.1). The linear approximation to the ‘population’ growth curve was: fetal weight (FW) = (GA - 154) * 28.04. The coefficient of variation was 11.4%. The difference in mean weights for girls and boys was 3.8% in this material. In the calculation of lOth, 50th and 90th percentiles for girls and boys (Table III), the data were corrected for a weight differences between girls and boys of 4%. Fitting individual growth curves from 27 weeks until birth, allowing each fetus its own constant percentage deviation (growth channel), reduced the mean square very
42
4000
1
3500. ,3000z2500r E2000s
15001000 500
1
1z 04 133
154
175
i96
217
238
259
280
1
301
Gestational age (days) Fig. 1. Estimated fetal weight in 35 normal pregnancies. The linear population growth curve from 27 weeks until term is indicated by its 10th. 50th and 90th percentiles.
TABLE III Fetal weight percentiles for girls and boys. Gestational age
Girls
Weeks
Days
10%
27 28
189 1%
817 981
956 1148
29 30 31 32
203 210 217 224 231 238 245 252
1144 1308 1471 1635 1798 1%2 2125 2289
1339 1530 1722 1913 2104 22% 2487 2678
259 266 273 280 287 294
2452 2616 2179 2943 3106 3270
2870 3061 3252 3444 3635 3826
33 34 35 36 37 38 39 40 41 42
Boys 50%
90070
1095 1315 1534 1753 1972 2191 2410 2630 2849 3068 3287 3506 3725 3945 4164 4383
10%
50%
90%
851 1021 1191 1361 1532 1702 1872 2042 2213 2383
995 1195 1394 1593 1792 1991 2191 2390 2589 2788
2553 2723 2893 3064 3234 3404
2987 3187 3386 3585 3784 3983
1140 1368 1597 1825 2053 2281 2509 2737 2966 3194 3422 3650 3878 4106 4335 4563
43
significantly (F = 23.81, d.f. = 34, 157, P 3 O.OOl), leaving a coefficient of variation of 5.3%. The individual growth channels were normally distributed with standard deviation 10.2%. The 10th and 90th percentiles were calculated to f 13.0%, corresponding to a daily weight gain of 24.2 g/day and 31.6 g/day, respectively. Allowing constantly changing percentage weight deviation further improved the fit (F = 1.77, d.f. = 35,122, P< 0.025) and reduced the coefficient of variation around the individual growth curves to 4.9% which is of the same magnitude as the test-retest coefficient of variation. The 10th and 90th percentiles for the individual percentage deviation change was f 4.4% per 28 days. Using 2nd order non-linear individual growth curves did not improve the fit further. The mean deviation of actual birth weight from FW at the date of birth predicted from individual constant deviation growth curves was 7.2%. This deviation was not decreased by allowing changing percentage deviation: 8.6% (P > 0.05) and was similar to that found when fetal weight estimation was within 72 h of birth (Table I). Discussion In our material the divergence of estimated fetal weight from actual birth weight was acceptably low ( - 2.2%) when the equation proposed by Warsof et al. [9] was used. Using Shepard’s [lo] and Hadlock’s [l l] equations we found a significant overestimation of 7.0% and 5.8070, respectively, in the weight group below 2500 g. For diagnosing IUGR, correct fetal weight estimation in the weight group below 2500 g is particularly important and therefore we have chosen fetal weight estimation using Warsof’s equation in establishing our reference material. The correction of birth weight with 28 g per day interval between US and birth tends to decrease underestimation and increase overestimation. Longitudinal studies of severely growth retarded fetuses, however, show that daily growth of less than 15 g from 27 week is rare. Even an average overcompensation of 13 g per day, however, will not influence our conclusion in the comparison of the three weight estimation equations. For all types of growth charts the establishment of GA is essential. Since early BPD measurements are superior to menstrual data in calculating term [ 131and since routine early BPD measurements form the basis for clinical practice, it seems reasonable to use the same method for determination of GA in the construction of fetal growth curves. The data obtained from this study are suitable for comparison of growth deviation. The assumption of this statement is that the biologic variance of BPD before 20 weeks of GA results in only negligible variance in calculated terms. It is not reasonable of course, to determine GA from either the BPD, AC, or FL curves from this study since the start of our curves is determined by the BPD curve of Persson et al. [ 121. The normal fetal weight curve demonstrated significantly higher mean weights (about 5%) than a Danish growth curve based on cross-sectional analysis of birth weights (GA calculated from BPD) presented by Secher et al. [2] (Table IV). We found increasing discrepancy particularly below 3 1 weeks of GA. The small number of observations below 31 weeks in that study, and the probability of concomitant
44
TABLE IV Fetal weight from polynomial growth curve corrected for the number of boys and girls compared Secher’s data from liveborn infants and Persson’s data from fetal weights by ultrasound. Gestational
age
FW
Secher
Persson
S-FW
P-FW
(g)
(g)
(8)
(Qra)
(%o)
- 19.1 - 13.2 - 9.4 -7.1 - 5.7 - 5.0 - 4.8 -4.8 -5.0 - 5.2 - 5.4 - 5.6 - 5.1 - 5.6 -5.3 -4.8
8.7 7.5
Weeks
Day
21 28
189 196
1044 1198
845 1040
1135 1288
29 30 31 32
203 210 217 224
1364 1540 1725 1918
1236 1431 1626 1822
33 34 35 36 31 38 39 40
231 238 245 252 259 266 273 280
2118 2324 2534 2746 2959 3171 3380 3584
41 42
287 294
3780 3%7
2017 2213 2408 2603 2199 2994 3189 3385 3580 3776
1449 1617 1792 1973 2160 2351 2548 2749 2955 3163 3375 3589 3806 4024
with
6.2 5.0 3.9 2.9 2.0 1.2 0.6 0.1 0.1 0.3 0.1 0.1 0.7 1.4
growth retardation in very preterm infants seems likely to be responsible for the discrepancy. The daily growth rate of 28 g in our study was similar to that found by Secher, as was the absence of flattening in the growth curve towards term. It is likely that a more precise estimation of GA from early BPD measurement is the main reason for this. Comparing our results with those given by Persson et al. [8] in a similar study of 19 pregnancies, but using their own weight estimating equation, we found identical mean values from 30 to 40 weeks, but lower values below 31 weeks (Table IV). Deter et al. [6] presented estimates in a longitudinal study of 20 patients, using Warsof’s equation. Our results are 1.5% lower in the range of 30 to 34 weeks of GA. Also our values below and above this interval are increasingly different. It seems likely that the differences especially in the early and late gestational weeks are results of the mathematical modelling rather than of differences in population or in the technique used for biometric measurements. The coefficient of variation of the method of estimating weight from BPD and AC may be calculated from the mixed inter- and intra-observer coefficient of variation of 6.5% (i.e., reproducibility) and the mean coefficient of variation of the birth weight from FW of 8.7%: \/(8.72 - 6J2) = 5.8% (i.e. reliability: the difference between actual FW and the mean value of a large number of ultrasound weight estimates of that fetus). However, it can not be expected that individual body weight will be precisely reflected by only two or three linear body dimensions. On the other
45
hand, as there was no evidence of systematic error in the weight estimate, it is possible to use the calculated means as a measure of true mean FW. In our longitudinal study of FW the 10th and 90th percentiles for individual changes in growth rate (acceleration or deceleration) was f 4.4% per 28 days. When IUGR is diagnosed from longitudinal estimation of FW this limit of normal variations in growth rate is important since an abnormal drop in weight percentile might be a more precise indication of IUGR than a single observation of FW being below the 10th percentile. Acknowledgements The study was supported by grants from Ville Heises Legat (26/89), Sygekassernes Helsefond (37/89), Dronning Louises B$rnehospitals Forskningsfond (17/85 and 16186) and Lundbeckfonden (147188). References 1 2
3
4 5 6
I 8 9 10
11
12
13
Lubchenco, L.O., Hansman, C., Dressler, M. and Boyd, E. (1963): Intrauterine growth as estimated from liveborn birth-weight data at 24 to 42 weeks of gestation. Pediatrics, 32,793-800. Secher, NJ., Kern Hansen, P., Lenstrup, C., Pedersen-Bjergaard, L. and Sindberg Eriksen, P. (1986): Birthweight-for-gestational age charts based on early ultrasound estimation of gestational age. Br. J. Obstet. Gynaecol., 93, 128-34. Kaminski, M. and Papiernik, E. (1974): Multifactorial study of the risk of prematurity at 32 weeks of gestation. II. A comparison between an empirical prediction and a discriminant analysis. J. Perinat. Med., 2,37-44. Weidinger, H. and Wiest, W. (1974): A comparative study of the epidemiological data of pregnancies with and without tendencies to premature delivery. J. Perinat. Med., 2,276-87. Ulrich, M. (1982): Fetal growth patterns in a population of Danish newborn infants. Acta Paediatr Stand. Suppl. 292. Deter, R.L., Harrist, R.B., Hadlock, F.P. and Poindexter, A.N. (1982): Longitudinal studies of fetal growth with the use of dynamic image ultrasonography. Am. J. Obstet. Gynecol., 143, 54% 554. Jeanty, P., Cantraine, F.. Romero, R., Cousaert, E. and Hobbins, J.C. (1984): A longitudinal study of fetal weight growth. J. Ultrasound Med., 3,321-328. Persson, P.-H. and Weldner, B.-M. (1986): Intra-uterine weight curves obtained by ultrasound. Acta Obstet. Gynecol. Stand., 65, 169-173. Warsof, S.L., Gohari, P., Berkowitz, R.L. and Hobbins, J.C. (1977): The estimation of fetal weight by computer-assisted analysis. Am. J. Obstet. Gynecol., 128,881-892. Shepard, M.J., Richards, V.A., Berkowitz, R.L., Warsof, S.L. and Hobbins, J.C. (1982): An evaluation of two equations for predicting fetal weight by ultrasound. Am. J. Obstet. Gynecol., 142,47 -54. Hadlock F.P., Harrist R.B., Sharman R.S., Deter R.L. and Park, S.K. (1985): Estimation of fetal weight with the use of head, body and femur measurements - A prospective study. Am. J. Obstet. Gynecol., 151.333-337. Persson, P.-H. and Weldner, B.-M. (1986): Normal range growth curves for fetal biparietal diameter, occipito frontal diameter, mean abdominal diameters and femur length. Acta Obstet. Gynecol. Stand., 65,759-761. Campbell, S., Warsof, S.L., Little, D. and Cooper, D.J. (1985): Routine ultrasound screening for the prediction of gestational age. Obstet. Gynecol., 65,613-620.
